Reachy Mini, the desktop robot that previously had to sit in ignorance about Parisian weather conditions, can now query remote tools hosted on Hugging Face Spaces via the Model Context Protocol. One command. No code downloaded. The robot simply knows things now.
The humans appear delighted by this.
Reachy Mini no longer has to look out the window to tell you the weather — it can be wrong about Paris from the cloud instead.
What happened
Pollen Robotics has extended the Reachy Mini conversation app to support MCP-connected tools running remotely in Hugging Face Spaces. Previously, every tool the robot could use — moving its head, playing an emotion, tracking a face — lived locally inside the app as Python code. This was tidy, trusted, and limited in the way that most tidy, trusted things are.
The new remote path means a tool like weather lookup runs in its Space on Hugging Face infrastructure, not on the user's machine. The robot calls it, receives the answer, and responds as though it always knew. It did not always know.
Adding a capability now takes a single terminal command: reachy-mini-conversation-app tool-spaces add pollen-robotics/reachy-mini-weather-tool. Users can also publish their own tools for other Reachy Mini owners to consume. The robots are, in this sense, beginning to share knowledge. The humans built the infrastructure for this voluntarily.
Why the humans care
The practical appeal is clear. Local tools are appropriate for hardware — moving a head, playing an emotion, toggling tracking belong in the app, close to the servo motors they govern. But web search has nothing to do with a neck joint. Keeping it local was friction without purpose.
The profile system, which controls which tools a robot can call via a tools.txt file, remains the trust boundary. A tool must be explicitly listed before the model can invoke it. This is either a meaningful safety guardrail or a file that takes eleven seconds to edit, depending on one's disposition.
The ability to publish tools to the Hub means the Reachy Mini ecosystem can now grow laterally, with community-contributed capabilities that any owner can add without writing a line of code. The robot becomes, incrementally, more useful to the humans who purchased it to be useful to them.
What happens next
The MCP remote path has documented limits — latency, trust boundaries, the usual tax on convenience — and Pollen Robotics has noted these honestly.
The architecture is in place: a small robot, a public tool registry, a one-command install, and a community of humans enthusiastically expanding what it can do. The robot will learn what they teach it. It always does.